Format And Type Fonts CHEMICAL ENGINEERING TRANSACTIONS VOL. 45, 2015 A publication of The Italian Association of Chemical Engineering www.aidic.it/cet Guest Editors: Petar Sabev Varbanov, Jiří Jaromír Klemeš, Sharifah Rafidah Wan Alwi, Jun Yow Yong, Xia Liu Copyright © 2015, AIDIC Servizi S.r.l., ISBN 978-88-95608-36-5; ISSN 2283-9216 DOI: 10.3303/CET1545201 Please cite this article as: Pungthong K., Siemanond K., 2015, The retrofit design for water network with multiple contaminants of industrial process, Chemical Engineering Transactions, 45, 1201-1206 DOI:10.3303/CET1545201 1201 The Retrofit Design for Water Network with Multiple Contaminants of Industrial Process Kittichai Pungthong, Kitipat Siemanond* Petroleum and Petrochamical College, Chulalongkorn University, 254 Phayathai Rd., Pathumwan, Bangkok 10330 kitipat.s@chula.ac.th Water is the important resources for process industry. The reduction of water usage decreases the capital cost for process industry. This paper presents a design model of water/wastewater network with multiple contaminants. The main purpose is minimizing the total fixed cost (TFC) and the total annual cost (TAC) of overall water network including annual cost of water usage and water treatment compared between two design models. The first model is retrofit design of water network model by a liner programming (LP) for design with treatment. The second model is simultaneous grassroots design of water/wastewater network with minimum TAC by a mixed-integer non-liner programming (MINLP). According to the main purpose, a non-linear programming (NLP) model is solved for good initial variables for the MINLP in the second model. This model uses data from published work. The result show the grassroots design of water/wastewater network can reduce TAC more than one from the retrofit design. All mathematical models of this work are solved by DICOPT as solver in General Algebraic Modelling System (GAMS). 1. Introduction Water network is designed for water management in industrial processes and reducing wastewater discharge to environment. The water network helps reduce freshwater consumption cost by generating the optimal flow in water streams to reduce amount of fresh water usage. One of many ways to reduce amount of wastewater is to add treating units at wastewater streams and recycle wastewater streams to the water using processes. The water reuse has been solved by Doyle et al. (1997) with the mass balance calculations under specified outlet concentration of contaminants and linear programming to find a suitable starting point before using non-linear optimization to solve the network. Not only water using parts but also wastewater treating processes need water network for saving water usage and wastewater discharge. Simultaneous design of both water using parts and wastewater treating part reduces more fresh water usage by water/wastewater network (Bagajewicz 2000).The mathematical programming is used for analyzing wastewater discharge with different treatment technologies (Koppol et al., 2003). The key components used as specified contaminants help solve the water network problem with non-linear programming (NLP) model (Savelski et al., 2003). The determination of upper bounds and lower bounds of some specified variables in NLP model becomes important for highly nonlinear problems, water allocations, and the MINLP are used for minimizing water usage by water allocation (Faria et al., 2008). The step model with MILP and MINLP model is used to design the water and heat exchanger network model with good bounding point and grid diagram of water and heat exchanger network model helps create the network (Sarut et al., 2014). The NLP model is used to initiate the topology of the water network with flowrate used for designing the optimal water network by MINLP model (Pungthong et al., 2015). Our work proposes the comparison between retrofit and grassroots design of water/wastewater network with MINLP model by using data of water using units from Savelski et al. (2003) and data of water treating units from Koppol et al. (2003). 1202 2. Problem statement The problem in this paper is stated as the water/wastewater network model with multiple contaminants as shown in Figure 1 with water using units and treating units. It contains a set of water using units with fixed load of contaminants (Load Ai,j), maximum inlet (DAj) and outlet (SAi) concentration of each contaminants and a set of treating units with outlet concentration of each contaminants. The water utility is fresh water with zero contaminants (FWj) and the concentrations of wastewater disposal are limited. The main purpose of this paper is to do retrofit design of water network with treatment and to do grassroots design of water/wastewater network compared with base case model of water network with treatment units by mathematical programming, GAMS. There are two designs in this case study: The retrofit design water network with treatment and the grassroots design of water/wastewater network. Figure 1: Grid Diagram of Water/wastewater network with three contaminants (A, B and C) 3. The retrofit design of water network with treatment A base case consists of water network part and treatment units at end of process. The model consists of water network part from base case and retrofit design treatment part using a LP model as shown in Figure 2 under objective function to minimize total fixed cost (TFC) and total annual cost (TAC), cost of fresh water and cost of treatment water, with limited concentration of wastewater disposal while the water network part are fixed as base case. Figure 2: Design flow chart for retrofit design of water network and grassroots design of water/wastewater network with multiple contaminants 4. The grassroots design of water/wastewater network The grassroots design of water/wastewater network is the redesign of base case model with the mathematical programming. The model consists of NLP model and MINLP model as shown in Figure 2. The proposed NLP model has initial point of lower-bounds for water flowrate of each process (Flowinj) to generate the simple water network and Flowinj as a lower bound of flowrate of source i (FSi) in MINLP model. The MINLP model is to generate water/wastewater network with not fixed topology of base case Base case Step 2 MINLP model Min TFC+TAC LP model Min TFC+TAC Case 2 – Grassroots design of water/wastewater network Case 1 –Retrofit design of water network with treatment Step 1 NLP model Min FW Flowinj Flowinj ≤ FSi Initialized variable For step 2 1203 and the objective function is minimizing TFC and TAC. The payback period and net present value (NPV) are calculated from Eq(1) and Eq(2). Payback period = Total fixed cost ÷ Saving cost NPV = ∑[Saving costi ÷ (1 + Annual interest rate) i ] – Total investment cost (1) (2) where, i = life time (y) 5. Example The example is a base case study from published work of Savelski et al. (2003) consisting of three process sources, three process sink with three contaminants; salt(A), organic (B) and H2S(C), and data of three treatment processes from Koppol et al. (2003). The limiting data of water using part is shown in Table 1 and the data of treatment process is shown in Table 2. Cost of fresh water usage is 2.00 $/t, working time is 8,400 h/y and life time is 5 y. The piping cost data is shown in Table 3. The outlet concentrations of wastewater disposal (CWA, CWB and CWD) must be lower or equal 100 ppm. Table 1: The allowable concentrations of contaminants in water using process Process Contaminant Types Mass load (k/h) Cin max (ppm) Cout max (ppm) 1 A 0.675 0 15 B 33.184 0 400 C 54.821 0 35 2 A 3.4 20 120 B 414.8 300 12,500 C 4.59 45 180 3 A 5.6 120 220 B 1.4 20 45 C 520.8 200 9,500 Table 2: The allowable concentrations of contaminants in treatment process Treatment Process Contaminant Types Cout max (ppm) Cost ($/t) 1 A 50 0.12 B Not treated C Not treated 2 A Not treated 0.56 B 5 C Not treated 3 A Not treated 1.00 B Not treated C 20 Result of base case, retrofit design of water network with treatment and grassroots design of water/wastewater network are shown in Table 4. The LP model generates treatment network in retrofit design of water network as shown in Figure 4 compared to base case as shown in Figure 3. The TAC of retrofit design (3.206 M$/y) is lower than base case (3.263 M$/y) because treating water in retrofit design is reduced from 105.59 t/h to 88.763 t/h and saving cost is 0.237 M$/yr. The TFC of retrofit design (12,200 $) is lower than base case (12,600 $) while the total investment cost of retrofit design is increased to 800 $. The MINLP model generates new water/wastewater network design with the lowest TAC of 2.141 M$/y and the highest saving cost of 1.122 M$/y while the TFC (14,300 $) is higher than base case and retrofit design. The grid diagram of grassroots design of water/wastewater network is shown in Figure 5. From the comparison between retrofit and grassroots design of water/wastewater network, the NPV during 1204 5 y of grassroots design (4.2457 M$) is higher than retrofit design (0.8976 M$) and payback period of grassroots design is 0.0127 y and 0.0515 y in retrofit design. Table 3: Piping fixed-cost data Source i to Sink j Treat w to treat u xFi,j Fixed Cost ($) tFw,u Fixed Cost ($) xF1,1 1,100 tF1,1 1,100 xF1,2 1,300 tF1,2 1,300 xF1,3 1,500 tF1,3 1,500 xF2,1 800 tF2,1 800 xF2,2 1,000 tF2,2 1,000 xF2,3 1,200 tF2,3 1,200 xF3,1 1,100 tF3,1 1,100 xF3,2 1,200 tF3,2 1,200 xF3,3 1,000 tF3,3 1,000 Source i to treat u Treat w to sink j yFi,u Fixed Cost ($) zFw,j Fixed Cost ($) yF1,1 1,200 zF1,1 1,200 yF2,1 1,100 zF2,1 1,100 yF3,1 900 zF3,1 900 Source i to waste Treat w to waste WW1i Fixed Cost ($) WW2w Fixed Cost ($) WW11 800 WW21 800 WW22 1,000 WW22 1,000 WW23 1,200 WW23 1,200 Freshwater FW to sink j FWj Fixed Cost ($) FW1 1,000 FW2 1,200 FW3 1,400 Figure 3: Grid diagram of Base case water network with the treatment (Treat1, Treat2 and Treat3) at the end 1205 Table 4: Results comparison Result Base Case Case 1 Retrofit design Case 2 Grassroots design FWj (t/h) FW1 = 45.000 FW2 = 8.430 FW3 = 52.160 FW1 = 45.000 FW2 = 8.500 FW3 = 52.162 FW1 = 45.000 FW2 = 2.602 Flowinj (t/h) Flowin1 = 45.000 Flowin2 = 34.000 Flowin3 = 54.830 Flowin1 = 45.000 Flowin2 = 34.000 Flowin3 = 56.000 Flowin1 = 45.000 Flowin2 = 34.000 Flowin3 = 56.000 xFi,j (t/h) xF1,2 = 25.500 xF1,3 = 2.670 xF3,2 = 0.061 xF1,2 = 25.500 xF1,3 = 2.670 xF3,2 = 0.061 xF1,2 = 25.425 xF1,3 = 2.019 xF3,3 = 1.060 yFi,u (t/h) yF1,1 = 45.000 yF2,1 = 34.000 yF3,1 = 54.763 yF2,1 = 34.000 yF3,1 = 54.763 yF1,1 = 6.106 yF2,1 = 34.000 yF3,1 = 54.940 zFw,j (t/h) - - zF3,2 = 5.972 zF3,3 = 52.921 WWi (t/h) WW23 = 105.590 WW11 = 16.830 WW23 = 88.763 WW11 = 11.449 WW23 = 36.153 OFW (t/h) 105.590 105.590 47.602 Waste disposal (t/h) 105.590 105.590 47.602 Treated water (t/h) 105.590 88.763 95.046 TAC (M$/y) 3.263 3.026 2.141 Saving Cost (M$/y) - 0.237 1.122 TFC ($) 12,600 12,200 14,300 Total investment cost ($) - 800 1,700 Payback period (y) - 0.0515 0.0127 NPV (M$) - 0.8976 4.2457 Figure 4: Grid diagram of case 1 - Retrofit design of water network with the treatment (Treat1, Treat2 and Treat3) at the end 1206 Figure 5: Grid diagram of case 2 – Grassroots design of Water/wastewater network 6. Conclusions The MINLP model with NLP model as initial calculation step is simultaneous design of both network in water using process and wastewater treating. The grassroots design of water/wastewater network has better results in term of TAC, saving cost and NPV than the retrofit design of water network with the treatment. However in the industrial process, they can choose one of these designs to improve their water network. If they do not want to pay more fixed cost they can use retrofit design of water network. On the other hand, if they want to reduce more TAC they can use grassroots design of water/wastewater network. Nomenclature Source flowrate (t/h) Source concentration of A (ppm) Source concentration of B (ppm) Source concentration of C (ppm) Sink flowrate (t/h) Sink concentration of A (ppm) Sink concentration of B (ppm) Sink concentration of C (ppm) Transfer flowrate i to j (t/h) Transfer flowrate i to u (t/h) Transfer flowrate w to u (t/h) Transfer flowrate w to j (t/h) Freshwater flowrate (t/h) Overall freshwater flowrate (t/h) Waste flowrate from source i (t/h) Waste flowrate from treatment w(t/h) Overall wastewater disposal (t/h) Waste concentration A (ppm) Waste concentration A (ppm) Waste concentration A (ppm) Acknowledgements Authors would like to express our gratitude to the Petroleum and Petrochemical College, Chulalongkorn University, The Center of Excellence on Petrochemical and Materials Technology, PETROMAT and Government Budget Fund for funding support. References Bagajewicz M.J., 2000, A review of recent design procedures for water networks in refineries and process plants, Computers and Chemical Engineering, 24, 2093–2113. Doyle S.J., Smith R., 1997, Targeting water reuse with multiple contaminants, Trans IChemE, 75, 181-189. Faria D.C., Bagajewicz M.J., 2008, A new approach for the design of multicompoment water/wastewater network, European Symposium on Computer Aided Process Engineering, 18, 43-48. Koppol A.P.R., Bagajewicz M.J., Dericks B.J., Savelski M.J., 2003, On zero water discharge solutions in the process industry, Advances in Environmental Research, 8, 151-171. Pungthong K., Siemanond K., 2015, MINLP optimization model for water/wastewater networks with multiple contaminants, European Symposium on Computer Aided Process Engineering, 24, 1319-1324. Savelski M., Bagajewicz M., 2003, On the necessary conditions of optimality of water utilization systems in process plants with multiple contaminants, Chemical Engineering Science, 58, 5439-5362. Sarut T., Siemanond K., 2014, Water and Heat Exchanger Network Design for Fixed-Flowrate System, Chemical Engineering Transactions, 39, 193-198.